基于线性神经网络处理圆度误差

圆度误差是机械零件及其互换性的重要指标,是产品质量的关键,这里提出一种基于线性神经网络计算圆度误差的方法.神经网络是一种非线性动力系统,具有运算和自适应的学习能力.本文首先分析线性神经网络的基本原理和实现方法,然后介绍利用神经网络进行圆度误差计算的详细方法,最后给出实验结果.仿真实验结果表明,本文介绍的方法可以有效、正确地评价圆度误差.该方法克服了利用最小二乘法评价圆度误差的局部收敛问题.

A Measurement Method of Roundness Error Based on Linear Neural Networks

The roundness error is an important index of mechanical parts and their interchangeable-ability, it is the key to quality of product, too. A new method for calculating the roundness error based on linear neural networks has been proposed. Neural networks are a nonlinear dynamical systems,which have operational and adaptive learning ability.Firstly the basic principle of linear neural networks is analyzed, and its implementation method is given.Then a detailed method for calculating the roundness error based on linear neural networks has been introduced.Finally experiment results are given. These results have shown that this method in this paper can correctly and effectively evaluate roundness error. The method overcomes the local convergence in evaluation of roundness error based on least square method (LSM).